nano-banana-2
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill requires the
@runcomfy/clipackage to be installed via npm to interface with the model API. - [COMMAND_EXECUTION]: The skill executes the
runcomfycommand-line tool with user-provided arguments to perform image generation tasks. - [CREDENTIALS_UNSAFE]: The skill references the use of the
RUNCOMFY_TOKENenvironment variable and local storage of API tokens in~/.config/runcomfy/token.jsonfor authentication. These are standard practices for the intended service and do not represent a credential leak in the skill code itself. - [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests user prompts and potentially external URLs to be processed by the image generation model. The documentation explicitly acknowledges these risks and notes that the CLI uses JSON serialization to prevent local shell injection.
Audit Metadata